callhome what local AI tools send out by default, and the switch for each

the phone-home register

Ollama

What Ollama sends out of your computer by default, when, to whom, and the switch that stops it, version by version, with how we know.

The answer

By its source at v0.34.4, not run: at start Ollama asks ollama.com for a list of recommended models (again about every 4 hours) and for its cloud-model catalogue, each request carrying the current time and signed with the installation's key, so the installation's public key rides on each call; OLLAMA_NO_CLOUD=1 in the server's own environment stops both, and also turns off cloud models and web search; the source at v0.32.14 shows the same two fetches at start under the same switch, its catalogue request signed in the same way and its recommended-models request unsigned.

on the research hub: A short history of Ollama · Ollama or vLLM: which to run, and how

Builds and versions

not installed: source read at tag v0.34.4On by default?yes · read in codeMeasured here?no · read in codeDate checked (UTC)2026-10-10 to 2026-10-11
Version
0.34.4 · the source at tag v0.34.4 (every platform it builds)
Where the evidence came from
not run: the source applies to the Linux, macOS and Windows builds
On by default?
yes · read in code
Measured here?
no · read in code
Date checked (UTC)
2026-10-10 to 2026-10-11
Why we say so
the source at tag v0.34.4: the start-up fetch, its gate on the cloud switch and its signed request. The source at the tag; we did not check that the released build matches it.; the source at tag v0.34.4: the fetch, its interval, its signature and its gate on the switch. The source at the tag; we did not check that the released build matches it.
not installed: source read at tag v0.32.14On by default?yes · read in codeMeasured here?no · read in codeDate checked (UTC)2026-09-28 to 2026-10-11
Version
0.32.14 · the Linux x86-64 server
Where the evidence came from
not run: the source applies to the Linux x86-64 server build
On by default?
yes · read in code
Measured here?
no · read in code
Date checked (UTC)
2026-09-28 to 2026-10-11
Why we say so
the source at tag v0.32.14, read 2026-10-11: the start-up fetch, its gate on the cloud switch, its version header and its signed request. The source at the tag; we did not check that the released build matches it.; the source at tag v0.32.14, read 2026-09-28: the fetch, its 4-hour interval and its gate on the switch. The source at the tag; we did not check that the released build matches it.

Each call out

cloud catalogue · 0.34.4

Version · build
0.34.4 · the source at tag v0.34.4 (every platform it builds) · not run: the source applies to the Linux, macOS and Windows builds · not installed: source read at tag v0.34.4
Sends by default
yes · read in code
Kind
fetches
Trigger
on its own
When
The code at v0.34.4 fetches the catalogue of Ollama's cloud models at start, in the background; we did not run this version.
Where
vendor API · ollama.com
What it sends
a request for the list of cloud models, then one request per cloud model for its details, each with the current time in its query (request); Ollama's own version, in the X-Ollama-Client-Version header of each request (software); the installation's public key, the same on every call from one installation, and a signature over each request and its time made with the installation's private key (identifier)
The vendor's stated purpose
“Adds a server-side cache for full `/api/show` responses, aimed at making model capability and metadata lookup fast for launch flows.” — pull request #15967, which added the cache (merged 2026-05-05), read 2026-10-10
Grade and method
read · the source at tag v0.34.4: the start-up fetch, its gate on the cloud switch and its signed request. The source at the tag; we did not check that the released build matches it. · read 2026-10-11
Receipts
the gate on the cloud switch, then the list and the details fetched · the two requests · the signed request to ollama.com · Ollama's version set in a header on each catalogue request · the default address · the header's name, X-Ollama-Client-Version · how the request is signed: the time added to its query, then the method and the address with that time signed · what the signature carries: the installation's public key, then a signature made with its private key · Ollama history, 'Who can talk to it' a house page; the receipts are above

cloud catalogue · 0.32.14

Version · build
0.32.14 · the Linux x86-64 server · not run: the source applies to the Linux x86-64 server build · not installed: source read at tag v0.32.14
Sends by default
yes · read in code
Kind
fetches
Trigger
on its own
When
The code at v0.32.14 fetches the catalogue of Ollama's cloud models at start, in the background; we did not run this version.
Where
vendor API · ollama.com
What it sends
a request for the list of cloud models, then one request per cloud model for its details, each with the current time in its query (request); Ollama's own version, in the X-Ollama-Client-Version header of each request (software); the installation's public key, the same on every call from one installation, and a signature over each request and its time made with the installation's private key (identifier)
The vendor's stated purpose
“Adds a server-side cache for full `/api/show` responses, aimed at making model capability and metadata lookup fast for launch flows.” — pull request #15967, which added the cache (merged 2026-05-05), read 2026-10-11
Grade and method
read · the source at tag v0.32.14, read 2026-10-11: the start-up fetch, its gate on the cloud switch, its version header and its signed request. The source at the tag; we did not check that the released build matches it. · read 2026-10-11
Receipts
the server starts its caches as it starts · the recommended-models cache and the catalogue cache started together · the catalogue fetched in the background at start · the gate on the cloud switch, then the list and the details fetched · the two requests · the signed request to ollama.com, with Ollama's version in a header · the default address, the version header's name and the signing step · how the request is signed: the time added to its query, then the method and the address with that time signed · what the signature carries: the installation's public key, then a signature made with its private key · the gate reads the cloud switch · the switch's name · Ollama history, 'Who can talk to it' a house page; the receipts are above

The switch

OLLAMA_NO_CLOUD=1

Accepted values
1
What it is
environment variable of the server process
Where it must be set
in the server's own environment: a systemd service needs a drop-in, a container needs it on its run line (see Turn them off) · turn them off
Other spellings
"disable_ollama_cloud": true in .ollama/server.json under the home folder of the account the server runs as
Also turns off
Ollama's cloud models; its web search
What turning it off costs
you lose Ollama's cloud models and web search, per its FAQ
Proof
found in code and documented; we have not published a test of it
For the call
cloud catalogue · 0.34.4

OLLAMA_NO_CLOUD=1

Accepted values
1
What it is
environment variable of the server process
Where it must be set
in the server's own environment: a systemd service needs a drop-in, a container needs it on its run line (see Turn them off) · turn them off
Other spellings
"disable_ollama_cloud": true in .ollama/server.json under the home folder of the account the server runs as
Also turns off
Ollama's cloud models; its web search
What turning it off costs
you lose Ollama's cloud models and web search, per its FAQ
Proof
found in code and documented; we have not published a test of it
For the call
recommended models · 0.34.4

OLLAMA_NO_CLOUD=1

Accepted values
1
What it is
environment variable of the server process
Where it must be set
in the server's own environment: a systemd service needs a drop-in, a container needs it on its run line (see Turn them off) · turn them off
Other spellings
"disable_ollama_cloud": true in .ollama/server.json under the home folder of the account the server runs as
Also turns off
Ollama's cloud models; its web search
What turning it off costs
you lose Ollama's cloud models and web search, per its FAQ
Proof
found in code and documented; we have not published a test of it
For the call
cloud catalogue · 0.32.14

OLLAMA_NO_CLOUD=1

Accepted values
1
What it is
environment variable of the server process
Where it must be set
in the server's own environment: a systemd service needs a drop-in, a container needs it on its run line (see Turn them off) · turn them off
Other spellings
"disable_ollama_cloud": true in .ollama/server.json under the home folder of the account the server runs as
Also turns off
Ollama's cloud models; its web search
What turning it off costs
you lose Ollama's cloud models and web search, per its FAQ
Proof
found in code and documented; we have not published a test of it
For the call
recommended models · 0.32.14

Listens beyond the machine, and other cloud features

Listens beyond the machine by default

  • loopback only · its server binds the loopback address by default (envconfig/config.go at v0.34.4); its official Docker image sets every interface (Dockerfile at v0.34.4), and its Docker guide publishes the port on the host; our page 'Ollama or vLLM: which to run, and how' (research hub, 2026-10-08), 'Who can reach it' (0.34.4 the source at tag v0.34.4 (every platform it builds))
  • loopback only · its server binds the loopback address by default (envconfig/config.go at v0.34.4); its official Docker image sets every interface (Dockerfile at v0.34.4), and its Docker guide publishes the port on the host; our page 'Ollama or vLLM: which to run, and how' (research hub, 2026-10-08), 'Who can reach it' (0.32.14 the Linux x86-64 server)

Other cloud features

  • cloud models (tags ending in -cloud) and web search, which run on ollama.com once you pick them · to avoid it: do not pick a cloud model; OLLAMA_NO_CLOUD=1 turns the features off · github.com · faq.mdx

What it leaves on disk

  • .ollama/cache/model-recommendations.json (under the server account's home) · the last list it fetched

Has it already run on your machine?

Brings in, and installed for you by

None on the register.

Credit

  • The switch and its file form are documented in the FAQ. — github.com · faq.mdx, read 2026-10-10
  • Ollama's FAQ says: "Ollama runs locally. We don't see your prompts or data when you run locally." The code read at v0.34.4 agrees for these two fetches: neither carries a prompt. — github.com · faq.mdx, read 2026-10-10
  • The switch and its file form are documented in the FAQ, and the server logs 'Ollama cloud disabled: true' when it takes. — github.com · faq.mdx, read 2026-10-10
  • Ollama's FAQ says: "Ollama runs locally. We don't see your prompts or data when you run locally." The code read at v0.32.14 agrees for these two fetches: neither carries a prompt. — github.com · faq.mdx, read 2026-10-10

Vendor documents it? · public discussion

Vendor documents it?

  • cloud catalogue · 0.34.4: partly · Its FAQ names the switch, the file form and what it turns off ('local only mode'), and says it does not see your prompts when you run locally; it does not describe the two fetches at start. · github.com · faq.mdx · read 2026-10-10
  • recommended models · 0.34.4: partly · Its FAQ names the switch, the file form and what it turns off ('local only mode'), and says it does not see your prompts when you run locally; it does not describe the two fetches at start. · github.com · faq.mdx · read 2026-10-10
  • cloud catalogue · 0.32.14: partly · Its FAQ names the switch, the file form and what it turns off ('local only mode'), and says it does not see your prompts when you run locally; it does not describe the two fetches at start. · github.com · faq.mdx · read 2026-10-10
  • recommended models · 0.32.14: partly · Its FAQ names the switch, the file form and what it turns off ('local only mode'), and says it does not see your prompts when you run locally; it does not describe the two fetches at start. · github.com · faq.mdx · read 2026-10-10

Public discussion

History

No dated changes yet.

Not yet checked

Nothing named.